01

Project context and disclosure boundary

The customer needed private LLM inference inside its own data center, keeping models, traffic, and operational logs within the enterprise network. The site had no dependency on public internet access, so model weights, Python packages, system installers, and container images had to be prepared and transferred offline.

To protect confidentiality, this case does not disclose the customer name, domain, personnel, server addresses, or internal security configuration. Only the anonymized architecture, delivery method, and test conclusions are presented.

02

Delivery challenges

Multi-node communication

Sixteen H200 GPUs were split across two inference nodes, requiring IB/RoCE validation and careful control of cross-node overhead.

Strict offline delivery

Weights, packages, and images had to be prepared externally, transferred by offline media, and backed by integrity evidence.

Large-model scheduling

Available memory could not host every target model simultaneously, so the design required one model online at a time, on-demand switching, and memory reclamation.

Operational readiness

The deliverable needed more than a running model: unified APIs, process supervision, recovery on boot, and repeatable test tooling were required.

03

Production architecture

Enterprise applicationsOpenAI-compatible API
Unified model gatewayLiteLLM · Routing and switching
Inference servicevLLM · Ray multi-node scheduling
Compute foundation2 nodes · 16×NVIDIA H200 · IB/RoCE
  • vLLM runs directly on the host to avoid unnecessary inference overhead.
  • Ray provides distributed scheduling across the two servers for GLM-5.2-FP8.
  • LiteLLM exposes a consistent OpenAI-compatible endpoint and isolates upstream clients from model switching.
  • Hermes Agent uses centralized deployment and separate user workspaces for controlled access.
04

Offline delivery and operations

01

Hardware and fabric validation

Validated GPUs, drivers, PCIe topology, IB/RoCE connectivity, and inter-node bandwidth.

02

Weight integrity

Calculated SHA256 checksums before and after transfer to verify offline migration integrity.

03

Service supervision

Used systemd for standardized start, stop, boot recovery, and automatic restart of Ray and vLLM.

04

Delivery evidence

Produced architecture, startup configuration, functional test evidence, and on-site operating instructions.

05

Functional verification

All six executed functional tests passed with no failures. An authorized fault-injection test remained disabled and is therefore excluded from this conclusion.

6 / 6Executed tests passed
0Failed tests
4 streamsConcurrent SSE verified
0 leakageIndependent sessions isolated
  • The model-list endpoint returned the intended model.
  • Non-streaming responses and completion states were correct.
  • SSE events were complete, ordered, and terminated normally.
  • Multi-turn context was retained without cross-session interference.
  • Four concurrent streams completed without request-ID confusion.
Scope of conclusion

These results represent functional verification, not final performance acceptance. Throughput, long-context behavior, concurrency tuning, and customer UAT remain subject to the formal acceptance report.

PLAN YOUR PRIVATE AI

Planning a multi-node private LLM cluster?

We size the model, context, concurrency, network, and security boundaries before defining on-site delivery and acceptance criteria.

Email us +86 139 2521 1225